The University of New South Wales 

Extraction of Building 

Outlines using LIDAR Imagery

Vassili Galettis 

Abstract

Various methods have been developed to measure the physical presence of objects with high positional accuracy. One of the most recent methods is the development of Airborne Laser Scanning, a method that scans the landscape through multiple passes, measuring the return time from pulses of laser light aboard the aircraft. The bi-product of the laser scan is a very accurate, dense cloud of points in 3D space that mirrors the landscape. The dilemma however is that in spite of the ability to measure very accurately the landscape in short period of time, the automatic detection and interpretation of the points remains difficult. Therefore the point clouds need to be modelled, so as to allow for the making of accurate 3 D maps.

Thus the aim of this thesis is to extract 3D GIS polygons of the tall buildings from raw LIDAR data, and produce a 3D UNSW campus model. A process achieved by Semi-Automatic Feature Extraction using the tools available within the ArcGIS suite.

What is LIDAR?

LIDAR is an acronym for Light Detection and Ranging, LIDAR technology uses Airborne Global Positioning with an Inertial Measurement Unit to determine the position and orientation of the system. The LIDAR unit uses a dispersed laser light pulsing at 25 kilohertz to make direct measurements from the aircraft to the earth's surface. The unit then measures distances to the first and last return from each pulse of light. The raw data represents a three-dimensional ‘First Surface Model’ and a ‘Bare-earth Model.’(AAM Geoscan. 2001) The bare-earth Model can be used to generate contours that represent the shape and elevation of the ground on accurate maps, and the combination of the two models can be used to estimate heights of features on the earth's surface. In this instance the buildings located on UNSW campus. The loaded data in last return DEM can be seen below in Qt modeler.

Extracting the Building Outlines 

Automated packages exist to extract building outlines however they are not readily available or reliable, as the algorithm is designed specifically for the data set. However tools available in ArcGIS can semi automate the process. 

By having the data in raster format the extraction can be done.

The photo below shows the extracted building outlines overlayed the last return UNSW campus model

To achieve this steps used include:

Disconnected Editing-> extract 

This allows the data the user  has selected to be separated from the main DEM

Spatial analyst -> Raster calculator

This allows the user to define the points of LIDAR data of which they want Example include using the Height greater than 45m but  less tha 50m.

ArcToolbox -> Conversion Tools -> From Raster -> Raster to Polygon

This converts to polygon so the buildings can be edited 

ArcToolbox -> Data management tools -> Features -> Features to Line

Here the user can get the general outline of the points selected

ArcToolbox -> Data management tools -> Generalization-> Simplified Line

The Line can now be simplified to fit the general shape of the building

ArcToolbox -> Data management tools -> Features -> Features to Polygon

The line now edited can be reclassified as a polygon so heights can be assigned to it to create a 3d model 

Assigning Heights and Slopes

Once the building outlines are extracted the building heights are assigned. Using the classify tool statistics can be obtained from the building height

Naturally all buildings do not have flat roofs. Thus the slope of the outline is analysed also

Here the spatial analyst tool can be used to determine the slope.

3D Model

Having the building outlines determined plus now the height and shape of the building all this information can be used to create a 3D model in ArcScene.

These buildings can the be overlayed a triangulated irregular network or TIN which is a 3D representation of contours of the UNSW campus

The final outcome

Results

The results were attempted to be of the highest possible accuracy the data would allow. In some of the cases, in particular the buildings, which were orthodox in shape, stood well above its surrounds or had limited noise, like vegetation surrounding their outline were extracted quite precisely. Examples of this include the Clancy auditorium as well as the UNSW Library.

However as with the major problem with the progression of LIDAR building extraction, overhanging vegetation, unorthodox building shape and clusters of buildings within close proximity, the results became worse, or couldn’t be extracted to accurate shapes

In these situations, automatic packages use buildings height and shapes are based on assumptions. This allows for the accuracy to be maintained too as high as possible, but cannot be guaranteed. The other significant difference is that the actual use of the LIDAR points does not occur. However depending on the scope of the work, these assumptions can be very appropriate. This is particularly true when there is a need for an estimation, approximation or rough guide. (Sithole G. 2005)

As extraction and LIDAR use is still relatively unexplored and incomplete, it was difficult to gauge the accuracy of this Thesis’ work. Obviously by keeping resolution at its highest, eliminating erroneous points, and limiting the noise surrounding the building all help, also a visual check that is to actually look at where the building outline lies in relation to the DEM data is also important, as this is a check for obvious gross errors.

Discussion

Advantages to Automation include:

Cost:  The approach above will only cost you a few man-hours and some processor time to explore if you have access to the Spatial Analyst extension. They will also allow you to make low cost assessments of the value of larger investments in extraction technology or projects.

Flexibility: You can focus your methods to your goals, your data and your expertise. Once you start to build your knowledge in the processes you will quickly be able to add or combine processes to meet new goals or needs, and you will be able modify the parameters of the process to suit your production and data environment. (Hewett M 2004)

Speed : Automated processes are generally faster than manual processes. At the very least, they can reduce a series of manual interactions to one simple command. In some cases, you can convert manual processes and assessments to fully automated processes.

Control : The user will be able develop processes specifically tailored to your data, and have the ability to make changes to processing procedures as needed to meet the challenges of your particular situation. The user will also have the ability to take processes you have developed and use them as modules of larger and more complex processes. Another aspect of control is the ability to apply the same settings to processes repeatedly without the possibility of typographic or other human error introducing inconsistencies between datasets. (Hewett M 2004)

3D models, particularly of residential and urban areas can provide new ways and alternatives for planners and developers to make informed decisions. The ability to model redevelopment areas represents one of the many applications of this technology. The models that are being developed will also further support necessary City services for land use planning, tax assessment, construction permitting, utility management, homeland security, and other major programs.( Pennington H 2005)  The integration of large datasets into these models will also allow emergency response technicians the ability make informed decisions in the field.

References

Sithole, G. (2005). Segmentation and Classification of Airborne Laser Scanner Data, Delft, The Netherlands

Hewett, M. (2004). Automating Feature Extraction with the ArcGIS Spatial. Analyst Extension. Document available online at

gis.esri.com/library/userconf/proc05/papers/pap2109.pdf

Author unknown. (2004). Automated building extraction and reconstruction using LIDAR data. Document available at

http://icrest.missouri.edu/Projects/NASA/FeatureExtraction-Buildings/Building%20Extraction.pdf.

Finlayson C. (2004). A complete Building extraction system from elevation data. Document available at

gis.esri.com/library/userconf/proc04/docs/pap1487.pdf

AAM GeoScan (2001). Scanning the Horizons, (17 Sep. 2001). Document available from AAMHatch on request

Further Information

For more information, please contact:

Dr S. LIm
Email: geomatic.eng@unsw.edu.au

Mail:
School of Surveying and Spatial Information Systems
University of New South Wales
UNSW SYDNEY  NSW  2052
Australia

Phone: +61-2-9385-4173
Fax: +61-2-9313-7493
WWW: http://www.gmat.unsw.edu.au